esa's climate change initiative and the aerosol...
TRANSCRIPT
ESA's Climate Change Initiative and the Aerosol ECVSimon PinnockESRIN17 May 2012 – ICAP Workshop
Aerosol CCI Team:
Acknowledgement to:
DLR Thomas Holzer-Popp, Dmytro Martynenko, Max Schwinger
FMI Gerrit de Leeuw, Larisa Sogacheva, Pekka Kolmonen
MPI-Met Stefan Kinne
Met.no Michael Schulz, Jan Griesfeller
U. Swansea Peter North, Andreas Heckel
U. Oxford Gareth Thomas, Don Grainger
RAL Richard Siddans, Caroline Poulsen
U. Bremen Wolfgang Hoyningen-Huene, Marco Vountas, Alex Kokhanovsky
HYGEOS Didier Ramon
LOA/ICARE Didier Tanre, Jacques Decscloitres, Pavel Litvinov, Francois-Marie Breon, Oleg Dubovic
BIRA Christine Bingen, Charles Robert
KNMI Deborah Stein-Zweers, Pepijn Veefkind
FUB Rene Preusker, Juergen Fischer
NILU Kerstin Stebel
PSI Paul Zieger, Urs Baltensperger
17 years of Aerosol Obs from ESA Satellites
– ESA doesn't have an instrument dedicated to aerosol retrieval flying yet.
(EarthCARE is coming soon ...)
– But information about aerosols is retrieved from several past instruments, often as a by-product of
the atmospheric correction, e.g.: MERIS, (A)ATSR, GOME, GOMOS, SCIAMACHY
– Continuity instruments will fly again on the future Sentinel-3A, 3B and 5P satellites, from 2014.
Apr 1995
Jun 2003
Mar 2002
Apr 2012ATSR-2
AATSR
MERIS
SCIAMACHY
GOMOS
ERS-2
ENVISAT
Sentinel-3A
SLSTR
OLCI
2014
TropOMISentinel-5P
EarthCARE
ADM-Aeolus
GOME
Sentinel-3B
SLSTR
OLCI
DUE GlobAerosol
Project Description
– User requirements: NWP, transboundary pollution, air quality
– Focussed on meeting users needs with what was already available
– GMV, U. Oxford, RAL, LOA, 2004-2009
– Production of a 12 year global aerosol dataset (1995-2007)
– ATSR-2, AATSR, MERIS, SEVIRI (slots: 10, 13, 16 UTC)
– Algorithms: Oxford-RAL "ORAC" retrieval + MERIS std L2 AOD
– Products:
– AOD, angstrom coeff, aerosol type
– 10 km and 1 degree resolution, global, daily, monthly, in netCDF with CF conv.
– Statistical fields (e.g. PDFs of AOD)
– Included per-pixel uncertainty estimates on AOD.
– Project ended with CTM model comparison/assimilation case studies: GEOS-Chem (U. Edinburgh),
GLOMAP (U. Leeds), IFS (ECMWF), LOTOS-EUROS (TNO), AEROCOM (MPI-M & LSCE)
– Data available from: www.globaerosol.info & ftp.globaerosol.info
DUE GlobAerosol
0
1
2
5
10
15
20
30
50
Num
ber o
f points in bin
0.55m Sea/ coastal
0.0 0.2 0.4 0.6 0.8 1.0Aeronet AOD
0.0
0.2
0.4
0.6
0.8
1.0
AATS
R
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Cor: 0.898Bias (y‐x): 0.0566Std.dev: 0.103Offset: ‐0.00610Gradient: 1.35
0.55m Land
0.0 0.2 0.4 0.6 0.8 1.0Aeronet AOD
0.0
0.2
0.4
0.6
0.8
1.0
AATS
R
0.0 0.2 0.4 0.6 0.8 1.0
0.0
0.2
0.4
0.6
0.8
1.0
Cor: 0.690Bias (y‐x): ‐0.0290Std.dev: 0.0708Offset: 0.0166Gradient: 0.608
0
1
2
5
10
15
20
30
50
Num
ber o
f points in bin
0.55m Sea/ coastal
0 1 2 3 4Aeronet Angstrom
0
1
2
3
4
AATS
R
0 1 2 3 4
0
1
2
3
4
Cor: 0.670Bias (y‐x): ‐0.285Std.dev: 0.400Offset: 0.279Gradient: 0.548
0.55m Land
0 1 2 3 4Aeronet Angstrom
0
1
2
3
4
AATS
R
0 1 2 3 4
0
1
2
3
4
Cor: ‐0.00893Bias (y‐x): ‐0.198Std.dev: 0.601Offset: 1.14Gradient: ‐0.00501
Oxford-RAL "ORAC" aerosol
G. Thomas et al., AMT, vol 2, 679, 2009
DUE GlobAerosol
Lessons Learned - 1:
1. AATSR AOD (ORAC) agreed with MODIS and MISR about as well as MODIS and MISR agreed with each other.
2. ESA Std MERIS L2 AOD is good for atm correction of MERIS data, but not ideal as an aerosol product.
3. (A)ATSR retrieval had room for improvement, although regional and seasonal patterns were quite well captured.
4. SEVIRI showed good potential over ocean.
5. Comparison of satellite-AOD with model-AOD was valuable in both directions, but satellite AOD considered to need
further improvement for routine model verification/assimilation.
Figure 1 Aerosol Optical Depth valid for 2005-08-04, 11:00-12:00 . Left: AOD observed by SEVIRI, retrievals from Globaerosol; averaged over model grid and hourly intervals. Middle: simulation from free running model. Right: assimilated field.
Arjo Segers, TNOLOTOS-EUROS
DUE GlobAerosol
Lessons Learned - 2:
1. Many algorithmic technical issues, such as:
– Desert dust and fire plumes often masked out by cloud clearing
– Retrieal over bright surfaces (desert, snow) needs to be improved.
– Regionally dependent positive biases and some of the variability in satellite AOD due poor cloud clearing.
– Angstrom coeff was poorly correlated with AERONET over land.
– Set of assumed aerosol optical properties insufficient to model reality.
2. Retrieving the aerosol type is really difficult (using retrieval cost function did not work)
3. Simple multi-satellite merged AOD product was not very useful.
4. Use of per-pixel uncertainties found to improve assimilation of SEVIRI AOD, compared to using fixed uncertainties.
5. Including the CTM model case studies was a good idea, as it resulted in very effective evaluation of the aerosol
products.
=> Development and Reprocessing cycle is necessary to build on lessons learned to improve the satellite
aerosol products:
Processing Evaluation Algorithm Improvement Reprocessing ... ...
GCOS Requirements for Satellite Observations
GCOS-82 in 2003GCOS-92 in 2004
GCOS-107 in 2006GCOS-138 in 2010
WORLD METEOROLOGICAL ORGANIZATION
INTERGOVERNMENTAL OCEANOGRAPHIC COMMISSION
SYSTEMATIC OBSERVATION REQUIREMENTS FOR SATELLITE-BASED DATA PRODUCTS FOR
CLIMATE
2011 Update
Supplemental details to the satellite-based component of the “Implementation Plan for the
Global Observing System for Climate in Support of the UNFCCC (2010 Update)”
December 2011
GCOS – 154
UNITED NATIONS ENVIRONMENT PROGRAMME
INTERNATIONAL COUNCIL FOR SCIENCE
GCOS-154 in 2011
GCOS Requirements for Aerosol
Systematic Observation Requirements for Satellite Based Data Products for ClimateGCOS-154, Dec 2011
CCI Programme Objective:
"To realize the full potential of the long-term global Earth Observation archives that ESA together with itsMember states have established over the last thirty years, as a significant and timely contribution to the ECVdatabases required by United Nations Framework Convention on Climate Change (UNFCCC)."
Proposed to ESA Ministerial Council in Nov 2008.Result: 6 Year Programme / 75 Meuro.
ESA's Climate Change Initiative
CCI Projects
DUE Projects
(GCOS-138)
14 CCI Projects www.esa-cci.org
cloud_cci DWD (D)
ozone_cci BIRA (B)
aerosol_cci DLR/FMI (D/FI)
ghg_cci U Bremen (D)
sst_cci U Edinburgh (UK)
land_cover_cci UCL (B)
sea_level_cc CLS (F)
ocean_colour_cci PML (UK)
glaciers_cci U. Zurich (CH)
fire _cci U.Alcala (E)
sea_ice_cci NERSC (N)
soil_moisture_cci TU Wien (A)
ice_sheet_cci DTU Space (DK)
CMUG UKMO - Hadley Centre (UK)
CCI Project Cardinal Requirements
1. Develop and validate algorithms to meet GCOS ECV requirements for (consistent, stable, error-characterized) global satellite data products from multi-sensor data archives
2. Produce, within an R&D context, the most complete and consistent possible time series of multi-sensor global satellite data products for climate research and modelling
3. Optimize impact of ESA EO missions data on climate data records
4. Generate complete specifications for an operational production system
5. Strengthen inter-disciplinary cooperation between international earth observation, climate research and modelling communities, in pursuit of scientific excellence
Not forgetting some CCI Principles: Transparency, open access to documentation and results, international collaboration, rigorous uncertainty characterisation
CCI Programme Overview
2010 2011 2012 2013 2014 2015 20162009
ITT
ITT
ITT
ITT
31 2co-location mid-term programme review
Phase 1 User Requirements Analysis Product Specification Algorithm Development Prototype Product Delivery Operational System Design
Phase 2 Continued Algorithm Development Operational System Implementation Operational Product Delivery
& Reprocessing
CMUG
Aerosol CCI Project
Phase 1: July 2010 – July 2013
www.esa-aerosol-cci.org
EO team
CMUGUK Met Office
ECMWFMeteo FranceMPI-Met
Science Leaders: Thomas Holzer-Popp (DLR) & Gerrit de Leeuw (FMI)
Approach
– Cyclic: user requirements development production validation user-evaluation
– Focus on common problems: e.g. surface reflectance, cloud-clearing, aerosol optical properties,
auxiliary data, product uncertainty specification, long term consistency, ...
– Investigate the relative strengths and weaknesses of different algorithms through intercomparison
on an equal footing.
– Perform idealised synthetic case studies
– Validation against AERONET, WMO-GAW
– Intercomparison with MODIS, MISR, POLDER, AEROCOM models, etc
– Evaluation by aerosol modellers and CMUG
– Develop an operational system to sustainably deliver aerosol ECV from satellite data
Aerosol CCI Project
[ iterate ]
Precursor Algorithms
– AATSR: ORAC (Oxford/RAL), ADV (FMI), Swansea (Swansea)
– MERIS: BAER (Bremen), ALAMO (HYGEOS), ESA Std MERIS AOD (LOV, LISE)
– AATSR+MERIS: Synergy (Swansea), SynAO (FUB)
– AATSR+SCIAMACHY: SYNAER (DLR)
– POLDER/PARASOL (LOA)
– OMI/AAI (KNMI)
– GOMOS/AERGOM (BIRA)
Deliverables (aim to meet GCOS requirements)
– Global 10km, 0.5deg, daily/monthly, AOD, angstrom, type*, absorption* in netCDF/CF-Conventions
– Single-sensor aerosol information from (A)ATSR, MERIS, PARASOL, OMI, GOMOS
– Synergy retrieval from AATSR+MERIS, AATSR+SCIAMACHY
Schedule
– First validated data sets for "golden year" 2008 will be available from Oct 2012
Aerosol CCI Project
Achievements since KO (26 July 2010)– Climate Requirements Review (GCOS, CMUG, AeroCom) and Product Specification
– Common format products from precursor algorithms (bigger job than it sounds!)
– Common aerosol properties defined and implemented (optical properties, size distributions and heights)
– types defined as mixtures of four simple components:
fine/weak abs., fine/strong abs., sea-salt and non-spherical dust
– component optical properties based on average properties measured by AERONET
– AeroCom median model merged with AERONET to define monthly climatologies of the distribution of
aerosol component fractions
Aerosol CCI Project
aerosol component
Refr. index, real part (55 m)
Refr. Index, imag part (.55 m)
reff ( m)
geom. st dev ( i)
variance (ln i)
mode. radius ( m)
comments aerosol layer height
Dust 1.56 0.0018 1.94 1.822 0.6 0.788 non-spherical
2-4km
sea salt 1.4 0 1.94 1.822 0.6 0.788 AOD threshold constraint
0-1 km
fine mode weak-abs
1.4 0.003 0.140 1.7 0.53 0.07 (ss-albedo at 0.55 m: 0.98)
0-2 km
fine mode strong-abs
1.5 0.040 0.140 1.7 0.53 0.07 (ss-albedo at 0.55 m: 0.802)
0-2 km
Achievements since KO – Part 2
Common cloud mask
– Probably biggest source of error, and particularly +ve AOD
bias over ocean
– Intercomparison of precursor cloud clearing on four 1-day
global fields and 17 test cases (incl. high aerosol dust and
smoke plumes and difficult cloud cases)
– Operational flags generally found to be unsuitable for aerosol
retrieval
– Selected APOLLO (AATSR), with addition of a dust flag and a
"safety zone"
– Post retrieval pixel quality control (threshold on # cloud free
pix per 10x10 km2 grid box)
Aerosol CCI Project
Achievements since KO – Part 3
– Lots of experimental algorithm variations tried out.
– Algorithm intercomparison and evaluation on 4 months of 2008 for competing AATSR and MERIS
algorithms, based on several experiments per algorithm – see next slide.
– Prototyping of AOD uncertainties for all algorithms
– Synthetic case studies:
– Extremely valuable feedback for the
algorithm developers
– General validation cannot be made
on a few idealised cases
– Definition of an operational Aerosol ECV
production system
– Algorithm devlopment (not covered further here):
OMI Aerosol Absorbing Index
GOMOS stratospheric aerosol
extinction profiles
Aerosol CCI Project
0.0 0.1 0.2 0.3 0.4 0.50.0
0.1
0.2
0.3
0.4
0.5
d=-0.05-0.15AOT
d=0.05+0.15AOT
550nm
AO
T(re
triev
ed)
AOT(true)
DLR Oxford-RAL Minsk FMI LOA-2 SRON Swansea
Product Validation and Algorithm Selection – Starting Point
Aerosol CCI Project
Product Validation and Algorithm Selection
Objective:
– Identify the best performing AOD retrieval (as well as best performing algorithm elements)
Protocol:
– Evaluations performed by independent partners assessing many different characteristics
– Data used for tuning retrievals was not used in validation
– Exercise was open to external participants
Intercomparison:
– Best performing algorithm variants were submitted to the round-robin.
– AATSR: FMI, Ox/RAL, Swansea, DLR/Synergy
– MERIS : ESA Std, HYGEOS/ALAMO (ocean only), BAER was not delivered in time.
– PARASOL std algorithm (fine mode only over land)
– Global, March, June, Sep, Dec 2008
– Reference data: AERONET, MODIS, MISR
– Did not consider the uncertainty estimates
Aerosol CCI Project
Product Validation and Algorithm Selection – Results
(Hot off the press, thanks to Stefan Kinne)
global ocean land– MISR v22 .62 .66 .59– MODIS aqua .55 .60 .50– MODIS terra .61 .63 .58– SEAWIFS .56 .58 .55
– AATSR F v13 -.57 -.60 -.55– AATSR S v30 -.46 -.48 -.48– AATSR O v11 .39 .40 -.39
PARASOL best ocean retrieval overallMERIS ALAMO (HYGEOS) better than ESA Std product over ocean. Ocean retrievals validated with coastal AERONET sites – did not use MAN data yet
Aerosol CCI Project
What do these scores mean ?• the more away from zero …. the better• the sign indicates the overall bias
direction• only involving regions with scores
MERIS not shown due to insufficiently significant statistics
Much lower coverage of AATSR compared to MODIS means that this comparison is biased to areas where MODIS performs poorly
Product Validation and Algorithm Selection – More Lessons Learned
– Team approach was successful to understand algo sensitivities and improve critical modules, resulting in clearly improved algorithms.
– Strong user involvement in the validation is essential
– More cycles of algorithm development and evaluation are needed
Last word on outcome from Stefan...
– "Current CCI retrievals for AOD have NOT (yet) reached the maturity of most US products"
– "ATSR (especially FMI) products are more competitive, but all CCI products lack coverage… usually less data than MISR"
Aerosol CCI Project
Where do we go from here?
Next 12 months (already funded)...– Production of one-year data sets from all candidate algorithms, followed by further validation and
intercomparison, delivery of first "official" validated Aerosol_CCI data set (Oct 2012).– Investigate and intercompare different approaches to handle surface reflectance (V.important over land)– Further work on cloud masking, synthetic case studies, etc– Work towards consistent cloud/aerosol products – collab. with Cloud_CCI project– Investigate humidity effects on aerosol optical properties.– Start development of AATSR+MERIS synergy algorithms in CCI– Processor perf. optimisation of new POLDER/PARASOL algorithm over land – aim to use as virtual AERONET site.
After (planned, but not yet funded)...– Production of full 17yr ATSR-2 + AATSR aerosol time series with best performing algorithm (requires work on
ATSR-2/AATSR overlap).– Production of full time series of MERIS, OMI, GOMOS, etc aerosol products– Intercomparison with other long time series aerosol CDRs (AVHRR, SeaWIFS, MODIS, MISR, TOMS, ...)– Development of an operational ECV production capacity
Operational System Specification
Or: How to implement "operational" production of climate data records.
Aerosol CCI Project
Expertise for validationand intercomparison
Developmentpriorities
Expertise foralgorithms
Validation / QCIntercomparisonStandard tools
Satelliteinput
Ancilliarydata
ArchiveDisseminationDocumentation
Expertise forECV use
Algorithmdevelopment
The Aerosol_cci system
Operational ECV Production system
External sciencecommunities
Spaceagencies
instrumentteams
Science Team
ExternalECV users
Referencedata
AerosolECV productsLv2 and lv3
Expertise for validationand intercomparison
Developmentpriorities
Expertise foralgorithms
Validation / QCIntercomparisonStandard tools
Satelliteinput
Ancilliarydata
ArchiveDisseminationDocumentation
Expertise forECV use
Algorithmdevelopment
The Aerosol_cci system
Operational ECV Production system
External sciencecommunities
Spaceagencies
instrumentteams
Science Team
ExternalECV users
Referencedata
AerosolECV productsLv2 and lv3
Some Questions for ICAP:
Q1: Aerosol_CCI
– Does anything need to be done to help NWP (i.e. ICAP) community benefit from work done in
Aerosol_CCI for climate?
e.g. Conversion to BUFR format, NRT production, model integration tools, assimilation experiments,
specific bias corrections, specific intercomparisons, ...
Q2: Sentinel-3
What needs to be done to maximise the ICAP community benefit from Sentinel-3 ?
– OLCI & SLSTR aerosol retrieval development (algorithms & validation) ?
– AOD data set production?
– Model integration tools?
Q3: Is there a need for a regular forum for satellite aerosol experts and modellers to meet?
– Does it exist already?
And finally...